Experiments with Automatic Vectorization For The Pentium(R) 4 Processor
Aart Bik, M. Girkar, P. M. Grey, Xinmin Tian, Intel Corporation, Microcomputer Software Labs
- Full Paper(.ps version)
The Pentium(R) 4 Processor features SIMD extensions to the Intel(R) Architecture that enhance the performance of computational intensive applications that perform the same operation on different elements in a data set. To take full advantage of these extensions, the high-performance Intel(R) C++/Fortran compiler developed at the Microcomputer Software Laboratories supports the automatic translation of serial loops into the SIMD instructions of the Pentium(R) 4 Processor. In this paper, we present an initial assessment of the effectiveness of this approach by studying the performance impact of automatic vectorization on some benchmarks.
Please contact our
webadmin with any comments or changes.
Unless explicitly stated otherwise, all material is
copyright © The University of Edinburgh.